Copilot's app and CLI now obey content exclusions, and an admin can set the default model

If your employer keeps certain files away from Copilot, that rule now follows you into the Copilot app and the Copilot CLI. GitHub made content exclusions generally available in both clients on September 2, 2026, for Copilot Business and Copilot Enterprise. The same day brought a second change: an administrator can name the model that new conversations start with.

What an exclusion actually stops

An exclusion is a list of paths written in fnmatch pattern notation, set by a repository administrator, an organization owner or an enterprise owner. GitHub's documentation lists four consequences for a file that matches a pattern.

A new rule does not apply the moment it is saved. The documentation says it can take up to 30 minutes for changes to take effect in environments where the settings are already loaded, so a rule written at the start of a meeting may not be in force by the end of it.

• Inline suggestions are not available in the affected files.
• The content of those files does not inform inline suggestions in other files.
• The content does not inform Copilot's responses in chat.
• The files are not reviewed in a Copilot code review.

Where the rule does not reach

An exclusion does not mean the model learns nothing about the file. The documentation says Copilot may use semantic information from an excluded file when the editor passes it along indirectly, and it names type information, hover-over definitions for symbols and general project properties as examples.

The second limit is easier to miss. Content exclusions do not apply to symbolic links or to repositories on remote filesystems. Anyone treating the feature as a guard on credentials or customer data should check how those files are reachable on the machine before relying on it.

The default model becomes an administrator's decision

The second change concerns model choice. In enterprise-managed settings the model key now takes the name of a specific model and version, and the documentation uses kimi-k-3 as its example. The change is generally available on Copilot Business and Copilot Enterprise, in the Copilot app, in Copilot CLI and in Visual Studio Code.

The setting governs only where a conversation starts. A user can still pick a different model inside a single conversation. An administrator who wants separate defaults per team marks the key overridable and maps teams through team-mappings.json, and a team that wants no imposed choice writes unmanaged in its own settings file.

Support is not uniform across clients. The documentation lists the model key as supported in Copilot CLI, Visual Studio Code and the GitHub Copilot app, and as unsupported in JetBrains IDEs and in the Copilot cloud agent.